Optimized Convolutional Neural Network Architecture for High-Accuracy Classification of Rose Leaf Diseases

Roses, a vital agricultural commodity in tropical and subtropical regions, face significant threats from various leaf diseases that reduce both yield and quality. This study focuses on improving rose leaf disease classification in Bangladesh by leveraging deep learning techniques. The proposed method utilizes a Convolutional Neural Network (CNN) to detect diseases such as rose rust and rose slug, which are common threats to rose cultivation. By training models on a dataset of 40,022 images containing healthy and diseased leaves, classifiers such as VGG-16, VGG-19, ResNet-50, and LCNN were employed to identify and classify diseases with an accuracy of 80%, 82%, 83%, and 82%, respectively. Preprocessing steps included data augmentation, cleaning, and normalization to ensure robustness in the model's performance. While image-based detection is highly effective, this research suggests that incorporating environmental and text-based data could further enhance disease detection accuracy and provide a more holistic approach to managing rose leaf diseases.